TwitterAnalysis
nsardar
2022-07-23
Loading the libraries
options(scipen = 999) #avoid the scientific notation
library(twitteR)
library(writexl)
library(bitops)
library(tidytext)
library(devtools)
library(sentiment)
library(quanteda)
library(quanteda.textstats)
library(quanteda.textplots)Getting Data
Based on the tweets from 4 different mobile manufacturing companies and then we will be doing text analysis and visualization for this companies.
We will be using twitter developer portal to get acces to tweets from twitter website.
consumer_key = "3N1fyHxwhVAVs5BCXzuQ8vbXV"
consumer_secret = "IVf6acHFxRWVQ006HNlDP9hygzOXx1XkZdKF9cMZlbL717AIQu"
access_token = "1458218683017347078-UTwl4ULAneM8wgDJypPXsU2KNhI8pU"
acess_token_secret = "4mLTYRLgJHSEoeV3qdz10WT7NdjwOH95AgvA8QCULhr0e"
setup_twitter_oauth(consumer_key,consumer_secret,access_token,acess_token_secret)## [1] "Using direct authentication"Mobile Companies
We will be using data from the below mentioned companies : 1) Motorola 2) Xiaomi 3) Oneplus 4) Samsung
tweetsMotorola = userTimeline('MotorolaUS', n = 1000)
tweetsXiaomi = userTimeline('Xiaomi', n = 1000)
tweetsOnePlus = userTimeline('oneplus', n = 1000)
tweetsSamsung = userTimeline('SamsungMobile', n = 1000)
tweetsDfMotorola = twListToDF(tweetsMotorola)
head(tweetsDfMotorola,2)## text
## 1 @JoeStep92374964 So sorry for the inconvenience this may have caused. Since you have mentioned that you have spoken… https://t.co/R17f6Bjvlq
## 2 @JPrice02 We're sorry for any inconvenience. If you wish to reconsider, we are standing by to help. -Erwin
## favorited favoriteCount replyToSN created truncated
## 1 FALSE 0 JoeStep92374964 2022-07-23 13:06:04 TRUE
## 2 FALSE 0 JPrice02 2022-07-23 00:51:55 FALSE
## replyToSID id replyToUID
## 1 1550718119001427968 1550829601890467841 1594103780
## 2 1550642166065659904 1550644844086820864 22788127
## statusSource screenName
## 1 <a href="https://www.sprinklr.com" rel="nofollow">Sprinklr</a> MotorolaUS
## 2 <a href="https://www.sprinklr.com" rel="nofollow">Sprinklr</a> MotorolaUS
## retweetCount isRetweet retweeted longitude latitude
## 1 0 FALSE FALSE NA NA
## 2 0 FALSE FALSE NA NAtweetsDfXiaomi = twListToDF(tweetsXiaomi)
head(tweetsDfXiaomi,2)## text
## 1 It is a beautiful thing to discover your own approach to style and hear your soul speak. #Xiaomi12Lite… https://t.co/L31RfFZybZ
## 2 Look exquisite, be confident, and stay determined with #Xiaomi12Lite at your side. #MyStyleMyShot https://t.co/uTVwCxkgtW
## favorited favoriteCount replyToSN created truncated replyToSID
## 1 FALSE 245 <NA> 2022-07-23 13:00:12 TRUE <NA>
## 2 FALSE 429 <NA> 2022-07-22 13:00:10 FALSE <NA>
## id replyToUID
## 1 1550828122404167680 <NA>
## 2 1550465728775704584 <NA>
## statusSource
## 1 <a href="https://prod1.sprinklr.com" rel="nofollow">Sprinklr Publishing</a>
## 2 <a href="https://prod1.sprinklr.com" rel="nofollow">Sprinklr Publishing</a>
## screenName retweetCount isRetweet retweeted longitude latitude
## 1 Xiaomi 25 FALSE FALSE NA NA
## 2 Xiaomi 38 FALSE FALSE NA NAtweetsDfOnePlus = twListToDF(tweetsOnePlus)
head(tweetsDfOnePlus,2)## text
## 1 Low light = blurry nights? Keep things sharp with the #OnePlus10T <U+0001F4F8>
## 2 What went into the #OnePlus10T's camera? Let's find out<U+0001F399><U+FE0F> https://t.co/8AJ7AF3Ojd
## favorited favoriteCount replyToSN created truncated replyToSID
## 1 FALSE 176 <NA> 2022-07-23 14:00:01 FALSE <NA>
## 2 FALSE 182 <NA> 2022-07-23 11:00:00 FALSE <NA>
## id replyToUID
## 1 1550843179389227010 <NA>
## 2 1550797874933276677 <NA>
## statusSource
## 1 <a href="https://twitter.com" rel="nofollow">Twitter for Advertisers</a>
## 2 <a href="https://mobile.twitter.com" rel="nofollow">Twitter Web App</a>
## screenName retweetCount isRetweet retweeted longitude latitude
## 1 oneplus 13 FALSE FALSE NA NA
## 2 oneplus 16 FALSE FALSE NA NAtweetsDfSamsung = twListToDF(tweetsSamsung)
head(tweetsDfSamsung,2)## text
## 1 Privacy is nothing without strong security. Your privacy. Secured. #SamsungKnox\n\nWhich feature makes you feel safe on your phone?
## 2 "What was once a novelty three years ago, is now the preferred choice for millions.” TM Roh, President & Head of MX… https://t.co/9CJngoKDK0
## favorited favoriteCount replyToSN created truncated replyToSID
## 1 FALSE 206 <NA> 2022-07-22 10:00:01 FALSE <NA>
## 2 FALSE 292 <NA> 2022-07-21 01:07:52 TRUE <NA>
## id replyToUID
## 1 1550420391751045122 <NA>
## 2 1549924084913688576 <NA>
## statusSource
## 1 <a href="https://bn.co" rel="nofollow">Brand Networks Composer</a>
## 2 <a href="https://studio.twitter.com" rel="nofollow">Twitter Media Studio</a>
## screenName retweetCount isRetweet retweeted longitude latitude
## 1 SamsungMobile 24 FALSE FALSE NA NA
## 2 SamsungMobile 31 FALSE FALSE NA NASentiment Analysis
The below score indicates whether the tweets are positive , negative or neutral
4 Positive
2 Neutral
0 Negative
finalExtractedTweets = rbind(tweetsDfMotorola, tweetsDfXiaomi, tweetsDfOnePlus, tweetsDfSamsung)
encodeSource <- function(x) {
if(x=="<a href=\"http://twitter.com/download/iphone\" rel=\"nofollow\">Twitter for iPhone</a>"){
gsub("<a href=\"http://twitter.com/download/iphone\" rel=\"nofollow\">Twitter for iPhone</a>", "iphone", x,fixed=TRUE)
}else if(x=="<a href=\"http://twitter.com/#!/download/ipad\" rel=\"nofollow\">Twitter for iPad</a>"){
gsub("<a href=\"http://twitter.com/#!/download/ipad\" rel=\"nofollow\">Twitter for iPad</a>","ipad",x,fixed=TRUE)
}else if(x=="<a href=\"http://twitter.com/download/android\" rel=\"nofollow\">Twitter for Android</a>"){
gsub("<a href=\"http://twitter.com/download/android\" rel=\"nofollow\">Twitter for Android</a>","android",x,fixed=TRUE)
} else if(x=="<a href=\"http://twitter.com\" rel=\"nofollow\">Twitter Web Client</a>"){
gsub("<a href=\"http://twitter.com\" rel=\"nofollow\">Twitter Web Client</a>","Web",x,fixed=TRUE)
} else if(x=="<a href=\"http://www.twitter.com\" rel=\"nofollow\">Twitter for Windows Phone</a>"){
gsub("<a href=\"http://www.twitter.com\" rel=\"nofollow\">Twitter for Windows Phone</a>","windows phone",x,fixed=TRUE)
}else if(x=="<a href=\"http://dlvr.it\" rel=\"nofollow\">dlvr.it</a>"){
gsub("<a href=\"http://dlvr.it\" rel=\"nofollow\">dlvr.it</a>","dlvr.it",x,fixed=TRUE)
}else if(x=="<a href=\"http://ifttt.com\" rel=\"nofollow\">IFTTT</a>"){
gsub("<a href=\"http://ifttt.com\" rel=\"nofollow\">IFTTT</a>","ifttt",x,fixed=TRUE)
}else if(x=="<a href=\"http://earthquaketrack.com\" rel=\"nofollow\">EarthquakeTrack.com</a>"){
gsub("<a href=\"http://earthquaketrack.com\" rel=\"nofollow\">EarthquakeTrack.com</a>","earthquaketrack",x,fixed=TRUE)
}else if(x=="<a href=\"http://www.didyoufeel.it/\" rel=\"nofollow\">Did You Feel It</a>"){
gsub("<a href=\"http://www.didyoufeel.it/\" rel=\"nofollow\">Did You Feel It</a>","did_you_feel_it",x,fixed=TRUE)
}else if(x=="<a href=\"http://www.mobeezio.com/apps/earthquake\" rel=\"nofollow\">Earthquake Mobile</a>"){
gsub("<a href=\"http://www.mobeezio.com/apps/earthquake\" rel=\"nofollow\">Earthquake Mobile</a>","earthquake_mobile",x,fixed=TRUE)
}else if(x=="<a href=\"http://www.facebook.com/twitter\" rel=\"nofollow\">Facebook</a>"){
gsub("<a href=\"http://www.facebook.com/twitter\" rel=\"nofollow\">Facebook</a>","facebook",x,fixed=TRUE)
}else {
"others"
}
}
finalExtractedTweets$tweetSource = sapply(finalExtractedTweets$statusSource,function(sourceSystem) encodeSource(sourceSystem))
tweetsSentiment<-sentiment(finalExtractedTweets$text) # instead used amazondf$text
finalExtractedTweets$polarity <- tweetsSentiment$polarity # Polarity denotes whther the tweet is neagtive, positive or neutral Final Results for Sentiment analysis on tweets
head(finalExtractedTweets,5)## text
## 1 @JoeStep92374964 So sorry for the inconvenience this may have caused. Since you have mentioned that you have spoken… https://t.co/R17f6Bjvlq
## 2 @JPrice02 We're sorry for any inconvenience. If you wish to reconsider, we are standing by to help. -Erwin
## 3 @JPrice02 Please let us know if you need help arranging the repair (2/2) -Erwin
## 4 @JPrice02 We understand. A swollen battery is not a safety issue. It is rare but typically occurs on older batterie… https://t.co/mUXwIF3DvD
## 5 @JPrice02 Hi, we're sorry to hear that. May ask if you are still having issues with your apps and gametime? Thanks -Erwin
## favorited favoriteCount replyToSN created truncated
## 1 FALSE 0 JoeStep92374964 2022-07-23 13:06:04 TRUE
## 2 FALSE 0 JPrice02 2022-07-23 00:51:55 FALSE
## 3 FALSE 0 JPrice02 2022-07-23 00:35:35 FALSE
## 4 FALSE 0 JPrice02 2022-07-23 00:35:07 TRUE
## 5 FALSE 0 JPrice02 2022-07-22 23:44:56 FALSE
## replyToSID id replyToUID
## 1 1550718119001427968 1550829601890467841 1594103780
## 2 1550642166065659904 1550644844086820864 22788127
## 3 1550633911218323457 1550640736571359233 22788127
## 4 1550633911218323457 1550640617901940736 22788127
## 5 1550620425583038464 1550627990064828418 22788127
## statusSource screenName
## 1 <a href="https://www.sprinklr.com" rel="nofollow">Sprinklr</a> MotorolaUS
## 2 <a href="https://www.sprinklr.com" rel="nofollow">Sprinklr</a> MotorolaUS
## 3 <a href="https://www.sprinklr.com" rel="nofollow">Sprinklr</a> MotorolaUS
## 4 <a href="https://www.sprinklr.com" rel="nofollow">Sprinklr</a> MotorolaUS
## 5 <a href="https://www.sprinklr.com" rel="nofollow">Sprinklr</a> MotorolaUS
## retweetCount isRetweet retweeted longitude latitude tweetSource polarity
## 1 0 FALSE FALSE NA NA others neutral
## 2 0 FALSE FALSE NA NA others neutral
## 3 0 FALSE FALSE NA NA others neutral
## 4 0 FALSE FALSE NA NA others negative
## 5 0 FALSE FALSE NA NA others positiveExtracting Results into excel to visualize in Tableau
#write_xlsx(finalExtractedTweets,'TweetsData.xlsx')Text Analysis and Visualization
We will be first visualizing text for individual companies and generate some isnights
ourCorpusMotorola = corpus(tweetsDfMotorola,text_field = 'text')
ourCorpusXiaomi = corpus(tweetsDfXiaomi,text_field = 'text')
ourCorpusOnePlus = corpus(tweetsDfOnePlus,text_field = 'text')
ourCorpusSamsung = corpus(tweetsDfSamsung,text_field = 'text')
# Lets remove few stop words and punctuation
ourTokensMotorola <- tokens(ourCorpusMotorola, remove_punct = TRUE)
ourTokensXiaomi <- tokens(ourCorpusXiaomi, remove_punct = TRUE)
ourTokensOnePlus <- tokens(ourCorpusOnePlus, remove_punct = TRUE)
ourTokensSamsung <- tokens(ourCorpusSamsung, remove_punct = TRUE)
ourTokensMotorolaNS <- tokens_select(ourTokensMotorola, stopwords('en'),selection = 'remove' , min_nchar =4)
ourTokensXiaomiNS <- tokens_select(ourTokensXiaomi, stopwords('en'),selection = 'remove' , min_nchar =4)
ourTokensOnePlusNS <- tokens_select(ourTokensOnePlus, stopwords('en'),selection = 'remove' , min_nchar =4)
ourTokensSamsungNS <- tokens_select(ourTokensSamsung, stopwords('en'),selection = 'remove' , min_nchar =4)
ourMatrixMotorola <- dfm(ourTokensMotorolaNS)
ourMatrixXiaomi <- dfm(ourTokensXiaomiNS)
ourMatrixOnePlus <- dfm(ourTokensOnePlusNS)
ourMatrixSamsung <- dfm(ourTokensSamsungNS)
# Top 5 features for each company
topfeatures(ourMatrixMotorola,5) # Motorola## thanks sorry reaching phone help
## 229 168 144 109 101topfeatures(ourMatrixXiaomi,5) # Xiaomi## xiaomi #risetothechallenge #mastereveryscene check
## 121 81 56 52
## #redminote11series
## 50topfeatures(ourMatrixOnePlus,5) # Oneplus ## #oneplus10pro oneplus @oneplus_in #shotononeplus find
## 61 45 32 30 16topfeatures(ourMatrixSamsung,5) # Samsung## #galaxys22 #samsungunpacked ultra series
## 289 245 159 139
## @bts_twt
## 99# Wordcloud for most used words in the 5 companies
textplot_wordcloud(ourMatrixMotorola,max_words = 200) # Motorola
textplot_wordcloud(ourMatrixXiaomi,max_words = 200) # Xiaomi
textplot_wordcloud(ourMatrixOnePlus,max_words = 200) # Oneplus
textplot_wordcloud(ourMatrixSamsung,max_words = 200) # Samsung
Let’s visualize plot for each company based on frequency of words which has been repeated atleast 10 times.
# Motorola
dfmIaugMoto = dfm(ourMatrixMotorola)
dfmIaugMoto = dfm_trim(dfmIaugMoto , min_termfreq=10)
textplot_wordcloud(dfmIaugMoto,max_words = 200 , min_size = 2)
# Xiaomi
dfmIaugXiaomi = dfm(ourMatrixXiaomi)
dfmIaugXiaomi = dfm_trim(dfmIaugXiaomi , min_termfreq=10)
textplot_wordcloud(dfmIaugXiaomi,max_words = 200 , min_size = 2)
# Oneplus
dfmIaugOnePlus = dfm(ourMatrixOnePlus)
dfmIaugOnePlus = dfm_trim(dfmIaugOnePlus , min_termfreq=10)
textplot_wordcloud(dfmIaugOnePlus,max_words = 200 , min_size = 2)
# Samsung
dfmIaugSamsung = dfm(ourMatrixSamsung)
dfmIaugSamsung = dfm_trim(dfmIaugSamsung , min_termfreq=10)
textplot_wordcloud(dfmIaugSamsung,max_words = 200 , min_size = 2)
Visualization of all the 5 twitter accounts together
our_corpus <- corpus(finalExtractedTweets,text_field = 'text')
head(our_corpus,2)## Corpus consisting of 2 documents and 17 docvars.
## text1 :
## "@JoeStep92374964 So sorry for the inconvenience this may hav..."
##
## text2 :
## "@JPrice02 We're sorry for any inconvenience. If you wish to ..."# Lets remove few words and punctuation
our_tokens <- tokens(our_corpus, remove_punct = TRUE)
words_to_be_removed<-c() # Currently we are not removing any additional extra words
our_tokens_nonstopwords<- tokens_select(our_tokens, pattern = c(stopwords('en'), words_to_be_removed),selection = 'remove', min_nchar = 4)
# Document feature Matrix to show frequency of tokens in each document
our_matrix<-dfm(our_tokens_nonstopwords)
our_matrix<-dfm_trim(our_matrix, min_termfreq = 10, min_docfreq = 5)
nfeat(our_matrix) #number of tokens in all documents## [1] 475#Most mentioned words
topfeatures(our_matrix,10)## #galaxys22 thanks #samsungunpacked sorry
## 289 274 245 169
## ultra like series reaching
## 162 154 151 150
## phone help
## 149 137topfeatures(our_matrix,30)## #galaxys22 thanks #samsungunpacked sorry
## 289 274 245 169
## ultra like series reaching
## 162 154 151 150
## phone help know xiaomi
## 149 137 129 121
## please hear check learn
## 117 117 112 108
## @bts_twt galaxy understand share
## 99 91 89 87
## thank time #risetothechallenge experience
## 85 84 81 78
## look best hello ready
## 75 75 74 74
## epic need
## 73 71set.seed(42)
textplot_wordcloud(our_matrix, max_words = 200, min_size = 1, max_size = 4) # All 5 companies combined
### Better Visualization of wordcloud for all the 5 companies
# Text plot showing visualization of all the companies together
corpus_subset(our_corpus, screenName %in% c("MotorolaUS", "Xiaomi", "oneplus","SamsungMobile")) %>%
tokens(remove_punct = TRUE) %>%
tokens_remove(stopwords("english")) %>%
dfm() %>%
dfm_group(groups = screenName) %>%
dfm_trim(min_termfreq = 5, verbose = FALSE) %>%
textplot_wordcloud(comparison = TRUE, max_words = 200)